TalentOS
Overview
TalentOS is a full-stack recruitment platform that combines authenticated workspace management, typed API contracts, and AI-assisted hiring signals. It demonstrates production-ready engineering patterns across React, Express, PostgreSQL, OpenAI integrations, CI, testing, Docker, and API contract generation.
Problem
Recruitment teams need structured tools for managing jobs, candidates, assessments, and hiring knowledge without turning AI into an ungrounded decision-maker.
Solution
Built a recruitment workspace with job and candidate management, structured candidate-to-job fit scoring, grounded RAG answers from workspace knowledge sources, typed OpenAPI contracts, and protected backend routes.
Architecture
React 19 and Vite frontend backed by an Express 5 TypeScript API. PostgreSQL with Drizzle ORM stores workspace data, Clerk handles authentication, OpenAPI 3.1 drives generated clients and schemas, and OpenAI integrations are isolated behind service-layer abstractions with Zod validation.
Technology stack
Screenshots
Key features
- Authenticated recruitment workspace with jobs, candidates, assessments, knowledge sources, and automations
- AI candidate evaluation with structured scores, strengths, gaps, and recommendation outputs
- Grounded RAG knowledge assistant using PostgreSQL full-text search and source-constrained answers
- OpenAPI 3.1 source of truth with generated React Query client and Zod schemas
- Rate limiting, security headers, request size limits, environment validation, and health checks
- 25 passing Vitest/Supertest tests covering API contracts, rate limiting, security, and readiness behavior
- Docker and Docker Compose support for repeatable development and deployment
Challenges
- Keeping AI features grounded and explainable instead of presenting them as autonomous hiring decisions
- Designing workspace-scoped data access and protected APIs across jobs, candidates, assessments, and knowledge sources
- Maintaining type safety from OpenAPI contracts through frontend clients and backend validation
Lessons learned
- AI product features need explicit boundaries, fallbacks, and source grounding to remain trustworthy
- Generated API clients and schemas reduce drift between frontend and backend contracts
- Security, testing, and deployment documentation make portfolio projects feel much closer to real products
Future improvements
- Document upload and parsing for knowledge sources
- pgvector embeddings for semantic retrieval
- Expanded automation execution engine
- Shared rate-limit store for horizontal scaling
- End-to-end test coverage